Killing COBOL in core banking systems

With 95% of ATM transactions relying on COBOL and approximately 800 billion lines of COBOL code estimated to be still in use, it’s clearly regarded as a reliable language, still able to power the world’s critical systems.

1: Thomson Reuters report ​
2: Microfocus report

While this is true, we’re heading rapidly into a time when COBOL is going to be present  significant challenges for most businesses. So, why haven’t we killed off COBOL? (And you might be asking, why do we need to kill off COBOL?!)

There are several factors at play here. It’sstill working but, much like the UK rail system, we’re not sure how much longer it can keep going. The average age of a COBOL developer is over 55, which means not only are we looking at a team of senior (expensive) engineers managing this language, but in the not too distant future we’refacing a skills gap epidemic when that average cohort reaches retirement age.  It’s also more than likely to be a long and tricky journey to replace your share of those 800 billion lines of code, so understandably it’s not something everyone is clamouring to get started on. The issue with these reasonable arguments to leave things as they are, is that COBOL is simply too risky not to replace.

This is the exact problem one of our customers faced. Picture a financial invoice processing system built on COBOL, responsible for processing €4.5 billion in revenue. The issue they faced was that this mission critical system was supported and maintained by only two COBOL developers, and one of those was planning to retire. This represented an existential risk to the business. If this system stopped working, they couldn’t process payments, simple as that. This issue was compounded by the fact that, as is typical of many systems, there was no documentation which explained the functional and technical behaviour.

Ultimately, we successfully transitioned them away from batch COBOL to a cloud native solution that derisked their business and catered to new organisational needs, introducing self-service functionality and enabling real time reporting and data driven insights.

However, as well as this project set out a template for how to successfully modernise a legacy COBOL solution, the harsh reality is that to derisk the delivery, we had to spend five months up front on documentation. 150 days paying a business analyst to document a system that had only 300,000 lines of code. Scale that up to systems that may contain many millions of lines of code and you can very quickly see that a lack of documentation, and the risk that introduces to the modernisation effort, very quickly puts that modernisation out of reach.

This is precisely why there is a struggle to kill COBOL. The traditional approach is just too slow.

But a real solution to all of this is here. Enter Generative AI. It’s already evident that this technology will have a transformative effect on software engineering and that the modernisation project of the future will be powered by GenAI. We had a vision that GenAI could be used to analyse, understand and automatically document legacy systems, thus derisking the modernisation process.

Version 1 created Decipher to do just that. AI powered code analysis and documentation. Automating the generation of functional and technical documentation to decode legacy codebases and enable efficient, cost-effective solutions.

Benefits of Decipher include:

• Reduce 20X+ investigation and documentation time
• Intuitive User Interface
• Comprehensive Technical & Functional Documentation
• Intelligent Summarisation​
• Ability to query code in natural language​

Analysing and creating documentation for an application with 1 million lines of code: ​

• With a mid-senior developer, this would take approximately 500 dev-days to complete
• With Decipher, the analysis and documentation runs and generates automatically in the background

Looking at a mid-range dev salary in the UK, averaging across geographies, this equates to a potential of around £100k in savings, minimum. When you think back to the fact that most of these COBOL devs are not mid-range, they’re seniors nearing the end of a successful career, that figure is realistically going to be significantly higher. Upwards of two or three times higher, and only set to increase.

Ultimately this solution removes a lot of the blockers currently in place for killing COBOL, which tips the scales directly in favour of removing that risk. To prepare for the future, you cannot rely on technology of the past. The game has changed, and companies are at higher risk than ever before of being left behind. The customer expectation now is for real time information, 24/7 services, global solutions, a cashless society, and more. All of this requires an unprecedented pace of innovation.

Let’s let those engineers retire in peace, kill COBOL, and set the standards for increasing customer demand, with a little help from generative AI.

And if you’re interested in the real world impact AI will have your modernisation and development activities, our next webinar takes place 20th August at 3pm, register here: AI-assisted  Application Modernisation